Almost every gold EA discussion ends with someone saying "the spread will kill it". That is true often enough to be worth quantifying instead of repeating. Here is the arithmetic, with the numbers I use when I evaluate a gold system.
The basic conversion
On standard XAUUSD contracts, one lot is 100 ounces. A price move of 0.01 (one point on a two-digit broker) is therefore one dollar per lot. On 0.01 lots it is one cent per point.
So a 30 point spread on 0.01 lots costs 30 cents per round turn: you pay it once when you enter, and if your exit is at market you effectively pay it again on the way out. If you want to be conservative, treat the round turn as two spreads, so 60 cents.
Now compare that to what a trade is expected to earn. If your average win before costs is 5 USD on 0.01 lots, a 60 cent round-turn is 12 percent of the gross win. That is survivable. If your average win is 1 USD, the same spread is 60 percent, and the system needs a very high win rate to exist.
Multiply by frequency
The number that matters is spread per hour, not spread per trade. A system taking two trades an hour at 30 points round turn on 0.01 lots spends 1.20 USD an hour. A system taking twelve trades an hour spends 7.20 USD an hour. Over a 20 hour trading day that is 24 USD versus 144 USD, which for a 1000 USD account is the difference between a normal month and a catastrophic one.
This is why the same strategy that works on M15 fails on M1 even when the logic is identical: the tax per hour went up six times while the edge per trade did not.
Widening spread regimes
Gold spreads are not constant. Around rollover, at market open, and during high impact news, the spread can widen by a factor of three to ten. If your backtest used a fixed 20 point spread and the live broker uses 60 points during the two hours when your EA wants to trade, the report is not wrong, it is just measuring a different market.
The practical approach is to run the test three times: once with the normal spread, once with the spread doubled, and once with the spread tripled during the specific hours your EA trades. If the system survives the third run, the spread is not the constraint.
The kill-shot scenario
The dangerous case is not a wider spread, it is a gap. Stops and stop-entry orders can fill far beyond their price when the market jumps. For gold, a 500 point slip is 5 USD on 0.01 lots, which is a full average trade's profit. Ten such fills in a month can erase the whole month.
That is why a news filter is not cosmetic. Avoiding the 30 minutes before and after a scheduled high impact release removes most of the gap risk for a price of missing a few trades.
What to measure on your own account
Export a month of fills and compute four numbers: average spread at entry, average spread at exit, average deviation between your intended price and the fill, and the worst ten deviations. Then divide total spread cost by gross profit. Under 20 percent is healthy. Between 20 and 40 percent means the strategy is living on borrowed time. Above 40 percent means you are paying the broker to trade.
Also check whether your broker charges commission per lot on top of the spread. On gold, some brokers charge 3 to 7 USD per lot per side. On 0.01 lots that is 6 to 14 cents round turn, which is small in absolute terms but adds another 10 to 20 percent to your transaction cost.
A short checklist
Write the spread into the strategy, not the report. Compute cost per trade, cost per hour and cost as a share of gross profit. Test with doubled and tripled spread. Prefer limit entries when the strategy allows it, because a limit order cannot slip against you. Use a news filter if the strategy trades breakout entries. And measure the live distribution monthly, because broker conditions change quietly.
None of this makes a bad strategy good. It simply stops you from discovering the cost after the account is already down 30 percent, which is how most gold EAs actually die.


